What's the status of Python 2.6 support?

Vincent Ladeuil v.ladeuil+lp at free.fr
Tue Dec 30 07:58:14 GMT 2008

>>>>> "jam" == John Arbash Meinel <john at arbash-meinel.com> writes:

    jam> Matt Nordhoff wrote:
    >> John Arbash Meinel wrote:
    >>> Matt Nordhoff wrote:
    >>>> Errm,
    >>>> Over the past few months, all known issues with Python 2.6 were fixed,
    >>>> right? 

All known issues raised by running the test suite have been
fixed, yes (except for a minor one I'm about to fix soon but it's

I keep running the full test suite from time to time, especially
before releases, but I've yet to automate that.

    >>>> But python2.6 hasn't been added to KNOWN_PYTHONS in the
    >>>> "bzr" script,

That's right.

    >>>> so bzr will only use 2.4 or 2.5.

That's not. If the 'python' found in the path is above 2.4, we
use it.

KNOWN_PYTHONS is used only if sys.version is lower than 2.4, so
if your default python is above that you're not concerned.

    >>>> Was that an oversight,

It was and is worth fixing (thanks for pointing it out), but
really it will trigger in pathological cases only (like in
museums where every python version is in the path and the default
one is lower than 2.4).

    >>>> or is there some reason not to use 2.6?

And even in the pathological cases, we will not use 2.6 if 2.4 or
2.5 is found first...

    >>> I'm pretty sure we should just add it to the list of

Do we really want to keep that code ? AIUI it makes sense only in
environments where the default python is 2.3.  I don't think
there are still older pythons around without a newer one
available and I doubt that environments can still propose
python-2.3 as the default python...

Can't we just get rid of the REINVOKE dance in favor of a simple
version check and leave distros handle the bang line ?

    >>> and make sure it starts getting tested properly.

That's a more important point than adding 2.6 to KNOWN_PYTHONS
IMHO ;-)

    >>> That would also include building 2.6 packages on
    >>> win32. I'm not sure if we want to default to py2.6 for
    >>> the standalone yet,

Having been surprised more than once by unexpected failures due
to 2.6, I think it make sense to wait for 2.6 availability on pqm
before making 2.6 the default (and even then, we may want to have
the test suite run regularly on windows first).


More information about the bazaar mailing list